ORCD will be teaching multiple classes to members of the MIT community during the summer of 2025 at MIT. See the class sections below for more details.
-
Course Description
The MIT Office of Research Computing's Engaging Cluster is available to the MIT community for running computational workloads that don't run well on your own computer. This hands-on tutorial walks you through the basics of using Engaging for your research. We will cover:
- How to access Engaging
- Transferring files
- Using and installing software
- Running jobs, including batch and interactive jobs requesting a variety of resources
Prerequisites/Requirements
- Attendees must have an Engaging account (instructions here).
- We ask that you read through our Getting Started Tutorial to familiarize yourself with the concepts beforehand.
- Attendees should also bring a laptop for the hands-on component.
Learning Objectives
After this tutorial attendees will
- Be able to log in and navigate the cluster
- Know how to use modules and have a plan for installing any additional software needed
- Be familiar with running jobs and requesting different types of resources
Schedule
This course is being run multiple times this summer; the same material will be covered in each instance of the class.
- June 25, 10AM-12PM
- July 30 2-4PM
- August 27, 2-4PM
Location
On-campus at ORCD's offices in NE36
How to Sign Up
Sign up for the June, July, or August class by selecting one of the dates at this Calendly link.
-
Course Description
Retrieval-augmented generation (RAG) is a framework for enhancing the knowledge capabilities of a pre-trained large language model (LLM) by providing it with a set of documents to use as a ground source of truth. RAG allows people to utilize the abstract reasoning and summarizing capabilities of LLMs to gain insights on the information provided in a given set of documents. In this workshop, we will learn how to run RAG using the GPUs on the Engaging cluster, as well as tailor the pipeline to work with any set of documents.
Prerequisites/Requirements
Required:
- Functional knowledge of Python
- Personal laptop
- Engaging account
- HuggingFace account
Recommended:
- Basic familiarity with containers (e.g., Docker, Apptainer)
- Basic understanding of LLMs
- Basic understanding of shell/bash commands
- Experience submitting jobs on an HPC system
Feel free to review the following before the workshop:
Learning Objectives
- Learning how to convert a set of documents into a vector store
- Learning how to implement a RAG pipeline using provided code and make changes based on individual needs
- Understanding how to use containers in an HPC setting
Schedule
- June 25, 2-4PM
Location
On-campus at ORCD's offices in NE36
How to Sign Up
To request the signup link, email the instructor, Sam Corey.
-
Course Description
Parallel computing has been an important research topic in science and technology for decades. Thanks to the fast-developing field of deep learning in recent years, parallel computing is being used for more broad interests. In this class, concepts of parallel computing will be introduced. Attendees will learn not only the basics of high-performance computing (HPC) clusters and GPU accelerators but also programming skills with OpenMP, MPI, CUDA, Pytorch, and Deepspeed. Examples and hands-on exercises will be provided in several programming languages including C, Fortran, and Python. These parallel programming skill sets are useful for researchers to accelerate their computer programs and helpful for students to be prepared for a career in information technology.
Prerequisites/Requirements
- Attendees should have minor programming experience in one of these languages: C, Fortran, Python, or Julia
- Attendees should also bring a laptop.
Learning Objectives
- Concepts of parallel computing and knowledge of accelerating computer programs on an HPC cluster.
- Parallel programming skills for CPU (OpenMP, MPI).
- Parallel programming skills for GPU (CUDA, Pytorch, Deepspeed).
Schedule
The topics on the two days are independent. Attendees can choose which session(s) they would like to attend.
- Day One (July 15):
- Parallel Programming with OpenMP: July 15, 10AM-12PM
- Distributed Computing with MPI: July 15, 1-3PM
- Day Two (July 16):
- GPU Programming with CUDA: July 16, 10AM-12PM
- Distributed Deep Learning: July 16, 1-3PM
Location
On-campus at ORCD's offices in NE36
How to Sign Up
To request the signup link, email the instructor, Shaohao Chen.